Tuesday, 13 October 2026 · Europe
EUR/USD 1.119 EUR/GBP 0.8446 EUR/CHF 0.9312 EUR/PLN 4.377 All rates →
Sign in · Join
EUROPES The European Report
European Edition Tuesday, 13 October 2026
LATEST How autonomous weapons are changing the way brigades fight Le Flash →
Tech & Startups

Google says Gemini 4 Argon is a ‘powerhouse’ AI model, and anyone can pay for it

Google says Gemini 4 Argon is a ‘powerhouse’ AI model, and anyone can pay for it

Google has released Gemini 4 Argon globally to paying users, in the Gemini app, in Search and through the API. The $2 and $10 per million token prices are a launch discount against standard rates of $4 and $20. Google has released Gemini 4 Argon to anyone willing to pay for it. The model is […] This story continues at The Next Web

Google has released Gemini 4 Argon globally to paying users, in the Gemini app, in Search and through the API. The $2 and $10 per million token prices are a launch discount against standard rates of $4 and $20.

Google has released Gemini 4 Argon to anyone willing to pay for it. The model is live globally in the Gemini app, in Search, and through the API.

Those prices are temporary. Argon’s standard rates are $4 and $20, and the launch discount runs for at least a month with no stated end date.

Tulsee Doshi, Gemini product lead at Google DeepMind, called it a “ powerhouse ” and the first of a family, she told Business Insider . Google says it ranks first on the Vals Index, which measures finance, coding, legal and tax work.

The first users were not developers. Argon went to a small group of cyber defenders without its guardrails , through a programme Google calls Fairwind.

That group is now growing to hundreds of partners. Wiz has been using the model to scan critical infrastructure, and says it found a bypass exposing a master key in healthcare software.

Europe has a name for this category. Cyber offence is one of four systemic risks the AI Office must police, alongside bio-attacks, loss of control and mass manipulation.

It polices them with 36 people. The office has had trouble getting access to some frontier models, and can fine up to 3% of global turnover when it does.

Google’s own numbers are reassuring, and they are Google’s. External bio red teaming unlocked under 1% of the sub-tasks it tests for, and shortcut attempts in coding fell to 2%.

It also runs a monitor that reads the model’s reasoning as it works, to catch a run going wrong before it finishes. Prompt injection attacks succeeded 0.7% of the time.

The internal gains come through Antigravity, Google’s agent tool. Agentic code submissions rose more than 35%, rollbacks fell by nearly a third, and a Chrome refactor took weeks instead of months.

The science claims are larger. Google says Argon helped find counterexamples to a 1967 conjecture about fusion reactor magnetic fields, and taught a humanoid to play cornhole by watching a video.

More from Tech & Startups